Extraction of object representations from stereo imagesequences utilizing statistical and deterministic regularities in visual data [1 citations — 0 self]
Abstract:
The human visual system is a highly interconnected machinery that aquires its stability through integration of information across modalities and time frames. This integration becomes possible by utilizing regularities in visual data, most importantly motion (especially rigid body motion) and statistical regularities re ected in Gestalt principles such as collinearity. In this paper we describe an articial vision system which extracts 3D{information from stereo sequences. This system uses deterministic and statistical regularities to aquire stable representations from unreliable submodalities such as stereo or edge detection. To make use of the above mentioned regularities we have to work within a complex machinery containing sub{modules such as stereo, pose estimation and an accumulation scheme. The interaction of these modules allows to use the statistical and deterministic regularities for feature disambiguation within a process of recurrent prdictions. 1
Citations
| 953 | Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography – Fischler, Bolles - 1981 |
| 56 | Integration of Visual Modules — An extension of the Marr Paradigm – Aloimonos, Shulman - 1989 |
